In vitro interactions between barley TALE homeodomain proteins suggest a role for protein–protein associations in the regulation of Knox gene function


For correspondence (fax +49 221 5062 413; e-mail


This paper describes two-hybrid interactions amongst barley homeodomain proteins encoded by the Three Amino acid Loop Extension (TALE) superfamily. The class I KNOX protein BKN3 is shown to homodimerise and to associate with proteins encoded by the class I and II Knox genes BKn-1 and BKn-7. Furthermore, JUBEL1 and JUBEL2, two BELL1 homologous proteins, are identified and characterised as interacting partners of BKN3. Differences in the requirements of BKN3 derivatives for interactions with KNOX and JUBEL proteins imply the involvement of overlapping but slightly different domains. This set of results is an example for interactions amongst different classes of plant TALE homeodomain proteins, as previously described for related animal proteins. Apparently identical spatial and temporal expression patterns of BKn-1, BKn-3, BKn-7, JuBel1 and JuBel2, as determined by in situ hybridisation, are compatible with possible interactions of their protein products in planta. Contradictory to the common model, that the transcriptional down-regulation of certain class 1 Knox-genes is the prerequisite for organ differentiation, transcripts of all five genes were, similar to Tkn1 and Tkn2/LeT6 of tomato, detected in incipient and immature leaves as well as in meristematic tissues. A characteristic phenotype is induced by the overexpression of JuBel2 in transgenic tobacco plants.


Although the first plant Three Amino acid Loop Extension (TALE) homeobox gene was discovered nine years ago (Vollbrecht et al., 1991), the knowledge of the role of this gene family in plant body organisation is relatively sparse (Bharathan et al., 1999; Pozzi et al., 1999; Reiser et al., 2000). TALE proteins differ from classical homeodomain proteins in the presence of three additional, conserved amino acids between helix 1 and helix 2 of the homeodomain (Bürglin, 1997). They are encoded by the class I and class II Knox genes and the Bell genes. Class I Knox genes are thought to play a role in the establishment and maintenance of meristematic identity, and in the initiation of leaf primordia. The dominant mutant alleles Kn-1, Rs-1, Lg-3, Gn-1 and TKn2/LeT6 cause alterations in leaf organisation (Foster et al., 1999; Hareven et al., 1996; Muehlbauer et al., 1999; Müller et al., 1995; Parnis et al., 1997; Schneeberger et al., 1995; Vollbrecht et al., 1991), and the overexpression of this group of genes has strong morphogenic effects in different plant species (Chan et al., 1998). Moreover, Arabidopsis plants defective in the class I Knox gene SHOOTMERISTEMLESS (STM) do not form a shoot apical meristem (SAM), supporting the role of this class I Knox gene in the establishment of meristematic identity (Long et al., 1996). The Bell genes represent a poorly studied second family of plant TALE genes. Loss-of-function of BELL-1 impairs the formation of the Arabidopsis ovule (Reiser et al., 1995), while the light-regulated ATH1 gene (Quaedvlieg et al., 1995) induces delayed flowering upon overexpression (Smeekens et al., 1998).

In Hooded barley, periclinal cell divisions in the primordium of the awn – a floral bract appendage – contribute to the formation of a meristematic cushion differentiating into an epiphyllous floret. This dominant phenotype is caused by a mutation in BKn-3 (Müller et al., 1995), the barley orthologue of the maize class I Knox gene Knotted-1 (Kn-1, Reiser et al., 2000). As in most dominant alleles of Kn-1, changes in a non-coding part of the wild-type gene are responsible for the Hooded phenotype (Müller et al., 1995). KNOX proteins contain several conserved regions N-terminal to the homeodomain: the KNOX and ELK domains and the GSE-box, which is absent in class II KNOX proteins. The KNOX domain and the closely related MEIS domain of certain vertebrate TALE proteins are probably derived from a common ancestral MEINOX domain (Bürglin, 1997).

The function of homeodomain proteins can be modulated by protein–protein interactions. By the association into complexes, transcription factors can be targeted to different DNA binding sites (Mann and Affolter, 1998). Here we report on protein–protein interactions of BKN3 with BKN1 and BKN7, encoded by barley class I and class II Knox genes, respectively (Müller, 1997). Furthermore, we describe the identification and characterisation of two novel BELL1-related barley TALE homeobox genes, encoding interacting partners of KNOX proteins.


Two-hybrid interactions of BKN3

Because the BKN3 protein fused to the GAL4 DNA-binding (BD) domain activated reportergene (HIS3 and LacZ) transcription in the two-hybrid yeast strain Y190, N-terminal deletion derivatives were tested for the absence of self-activation. A non-activating derivative corresponding to BKN3 amino acids 111–364, including the KNOX, GSE, ELK and homeodomains, was identified. The expression of the fusion protein in yeast was verified by Western-blot analysis (data not shown). In a similar approach, an N-terminal deletion derivative of the barley class I KNOX protein BKN1 (amino acids 77–349) and an N-terminally extended derivative of the class II KNOX protein BKN7 were identified as suitable for two-hybrid studies. None of the tested activation domain (AD) fusions caused reporter gene activation in the absence of an interacting BD fusion protein. Coexpression of any AD and BD fusions of BKN1, BKN3 and BKN7 induced the expression of both reporter genes in Y190, indicating the formation of homo-and heterodimers of the three KNOX proteins. Figure 1 illustrates a typical outcome of the interaction assays monitored by β-galactosidase activity. These experiments were repeated several times and confirmed by interaction-dependent growth on culture medium lacking histidine (data no shown).

Figure 1.

Yeast two-hybrid assays with barley KNOX proteins.

All combinations of AD and BD fusions of the three homeodomain proteins were tested for induction of β-galactosidase activity in the yeast strain Y190. Yeast strains carrying the appropriate constructs were grown to equal optical density in liquid culture and dot-blotted on a nitrocellulose filter. LacZ-assays were carried out according to Essers and Kunze (1996). None of the constructs supported reporter gene expression in the presence of the GAL4 AD or BD domain, respectively.

For the identification of further interacting partners of BKN3, 106 clones of a two-hybrid cDNA library constructed from young barley inflorescences were screened with the bait BKN3(111–364) for the activation of the HIS3 reporter gene. By this approach two different genes represented by four clones were identified. The interaction of the products of these clones with BKN3(111–364) was confirmed by retransformation and β-galactosidase tests. Both cDNA inserts contained a region with similarity to the putative coiled-coil domain of the BELL1 gene product from A. thaliana (Reiser et al., 1995). The two BELL1 homologues were designated JuBel1 and JuBel2.

The JuBel1 and JuBel2 genes

Entire open reading frames (ORFs) were obtained by the screening of a barley inflorescence cDNA-library and by 5′-RACE amplification of inflorescence mRNAs with gene-specific primers. The JuBel1 cDNA of 2960 bp encodes an ORF of 760 amino acids (Figure 2a). The JuBel2 transcript is 2130 nucleotides in length and contains an ORF of 608 amino acids (Figure 2b). The homeodomains of JUBEL1 (amino acids 530–592) and JUBEL2 (amino acids 343–405) show 85.5% amino acid identity to each other and 88.5 and 85.2% amino acid identity to BELL1 from Arabidopsis, respectively. The presence of the three conserved amino acids PYP between helix 1 and 2 of their homeodomains defines them as members of the TALE superclass of homeodomain proteins. A second region of similarity amongst BELL proteins is located at the N-terminal end of the homeodomain. It contains a putative coiled-coil motif spanning amino acids 402–440 in JUBEL1 and 217–264 in JUBEL2, including a region with similarity to a nuclear localisation signal (NLS; Reiser et al., 1995).

Figure 2.

cDNA and deduced amino acid sequences of JuBel1 and JuBel2.

(a) cDNA and amino acid sequence of JuBel1.

(b) cDNA and amino acid sequence of JuBel2.

The putative coiled-coil domains are shaded in grey, the putative nuclear localisation signals (NLS) are underlined. Black shading corresponds to the homeodomains. Arrowheads mark the positions of introns; asterisks indicate stop codons.

Genomic clones covering the JuBel1 and JuBel2 genes were obtained by the screening of a genomic barley library with PCR-amplified intron probes. Both genes carried three introns in identical positions of the translated regions with respect to the homologous genes from Arabidopsis(Figure 2a,b). RACE experiments suggested the existence of an additional intron in the non-translated leader of JuBel1. All but one intron were flanked by the typical exon/intron borders. The second intron of JuBel1 contains GC as a 5′ splice site; use of such a site has been found to occur at a frequency of 0.4–1% in plant introns (Simpson and Filipowicz, 1996). Intron sizes varied considerably (JuBel1: 278 bp, 1428 bp, 1284 bp and 4878 bp; JuBel2: 770 bp, 1366 bp and 99 bp). As determined by Southern hybridisation, JuBel1 and JuBel2 were present in the barley genome as single copies, but low-stringency hybridisations gave indications of further members of this gene family (data not shown).

The genomic JuBel1 sequence included the RFLP marker MWG2299, defining a polymorphic locus at position S 38–41 (52.4 cm from the start of the chromosome) of linkage group 4H in the Igri × Franka map (Michalek et al., 1999). The mapping of JuBel2 to the map of Castiglioni et al. (1998) was carried out by SNP heteroduplex analysis based on 100 random lines from a Proctor × Nudinka mapping population (Heun et al., 1991). JuBel2 mapped on chromosome 3H in the linkage subgroup 28 close to marker E42385 (Castiglioni et al., 1998).

In vivo and in vitro interactions of BKN1, BKN3, BKN7, JUBEL1 and JUBEL2

Non self-activating AD and BD fusions of BKN and JUBEL deletion derivatives were subjected to two-hybrid interaction assays. Semiquantitative β-galactosidase assays allowed a rough discrimination between strong and weak interactions.

BKN3 amino acid residues 1–167 were not necessary for its homodimerisation and for interactions with BKN1, BKN7 or JUBEL1 (Figure 3). The interaction with JUBEL2, however, required BKN3 residues between positions 111 and 167, when the complete C-terminus of the protein was present. Deletion of the BKN3 C-terminal amino acids 236–364 strongly reduced its ability to interact with the other KNOX proteins, while amino acid residues 111–236 were sufficient to ensure strong interactions with both JUBEL proteins. Surprisingly, the presence of BKN3 residues 277–308 abolished the interaction with JUBEL1, but not with JUBEL2. In summary, the C-terminal part of the KNOX domain (in the case of the BKN3–JUBEL2 interaction, the entire KNOX domain) and the region between the KNOX and ELK domains, including the GSE box, were sufficient for strong interactions with both JUBEL proteins and weak associations with BKN1 and BKN3. Strong KNOX–KNOX interactions required the complete C-terminal segment of BKN3.

Figure 3.

Two-hybrid interactions of different KNOX- and BELL deletion constructs measured by semiquantitative β-galactosidase assays.

The domain structure of the deletion derivatives is schematically indicated. Numbers in parentheses correspond to amino acid residues included in AD and BD fusion proteins.

b: basic domain; coiled-coil: putative coiled-coil; ELK: ELK domain; GSE: GSE box; HD: homeodomain; KNOX: KNOX domain. Weak (+) and strong interactions (+ +) were discriminated based on staining intensities in the -galactosidase assays. (+)/(+ +): interactions observed only with either the AD or the BD fusion of the corresponding peptide.

The KNOX-interacting domain of JUBEL1 was narrowed down to amino acids 285–423, including the putative coiled-coil domain. However, this conserved region alone (amino acids 329–391) was not sufficient for the interaction. JUBEL1 interactions with BKN7 were less clear-cut, as AD and BD fusions of some peptides showed different associations. The shortest tested derivative of JUBEL2 (amino acids 122–341) interacting with BKN1 and BKN3 also contained the putative coiled-coil domain. None of the JUBEL2 derivatives interacted with BKN7 in the two-hybrid assays. JUBEL1 and JUBEL2 formed neither homo- nor heterodimers.

Two-hybrid interactions amongst the BKN and JUBEL proteins were confirmed by in vitro pull-down assays. Derivatives of BKN1, BKN3, BKN7 and JUBEL1 were expressed as glutathione-S-transferase (GST) fusion proteins in E. coli, adsorbed onto glutathione-Sepharose and incubated with in vitro translated, radioactively labelled BKN3, JUBEL1 or JUBEL2. Interactions of labelled proteins with the GST fusions were monitored by SDS-PAGE and autoradiography. Labelled BKN3(1–364) clearly associated with the GST fusions of BKN1, BKN3, BKN7 and JUBEL1. In vitro translated JUBEL1(185–760) and JUBEL2(122–341) were also shown to interact with GST derivatives of BKN1 and BKN3. A weak association of both JUBEL peptides with GST-BKN7 was also observed, while no indications for interactions between JUBEL1 and JUBEL2 were found (Figure 4).

Figure 4.

In vitro interaction assays.

Interactions of the radioactively labelled in vitro translation products (asterisks) BKN3(1–364) (upper panel), JUBEL1(185–760) (middle panel) and JUBEL2(122–341) (lower panel) with GST fusions of BKN1(1–349), BKN3(1–364), BKN7(1–349) and JUBEL1(185–760). As negative controls in vitro translated proteins were incubated with glutathione-Sepharose (M) and glutathione-sepharose-coupled GST (GST).

Expression patterns of the BKn and JuBel genes

Spatial and temporal expression patterns of the barley Knox and JuBel genes were determined. By RT–PCR and Northern blot analysis, transcripts of all five genes could be detected at different levels in the embryo, the seedling root and shoot, stem tissue, very young leaves, inflorescences and floral organs (data not shown). Examples of in situ hybridisations are shown in Figure 5 (for a more detailed presentation see The expression domains of the five homeobox genes overlapped spatially and temporally, assuring that interactions amongst their translation products are topologically possible in planta. Transcripts were detected in meristematic tissues and young lateral organs at different stages of development. In the mature embryo, expression was observed throughout the SAM, including the L1-layer of the meristem (Figure 5f), and in leaf primordia, in the provascular tissue of the embryo axis and in the embryonic primary and adventitious roots. Signals were associated with the vascular strand connecting the embryonic axis with the scutellum and with the vascular strands of the scutellum and the coleoptile (Figure 5a–e). At the seedling stage, expression was detected throughout the SAM, in young leaves and their primordia, in the proximity of vascular strands of the stem and in the primary and adventitious roots (Figure 5g). During inflorescence development transcripts of all five homeobox genes were present in the inflorescence and floral meristems, in floral organs and their primordia and were associated with vascular strands (Figure 5h–j).

Figure 5.

Expression analysis of the BKn-1, BKn-3, BKn-7, JuBel1 and JuBel2 genes.

(a–e) Longitudinal sections of mature barley embryos (variety Atlas) hybridised with BKn-1-(a), BKn-3-(b), BKn-7-(c), JuBel1-(d) and JuBel2-(e) specific antisense probes (see Experimental procedures). The detection of BKn-7 signals demanded an approximately four-fold longer incubation period than the other genes.

(f–j) Hybridisations with a JuBel2-specific probe. (f) Close-up of the SAM of a mature embryo. (g) Longitudinal section of a seedling 4 d after germination. (h) Longitudinal section of a wild-type inflorescence at the lemma primordium stage. (i) Wild-type inflorescence at the awn primordium stage. (j)Hooded inflorescence at the white anther stage. In situ hybridisations with BKn-1, BKn-3, BKn-7 and JuBel1 specific probes produced indistinguishable expression patterns.

These results, including negative controls, can be accessed at Scale bars correspond to 200 m.

an: anther; co: coleoptile; cu: meristematic cushion on the Hooded lemma; gl: glume; le: lemma; lb: lateral bud; ov: ovary; P1, P2, P3: leaf primordia; pa: palea; r: root; sc: scutellum; sr: seminal root; va: vascular strands.

Overexpression of JuBel2 in transgenic tobacco

While the effect of the overexpression of JuBel1 in transgenic tobacco under control of the CaMV 35S promoter was restricted to a reduction of male fertility, the expression of a N-terminally truncated JUBEL2 derivative caused morphological alterations in transgenic tobacco. An impaired regeneration capacity of transgenic JuBel2 calli allowed the recovery of only 18 independent JuBel2-transgenic T0 plants. Six of them showed significant deviation from wild-type morphology, and their progenies were further analysed. In sterile culture, some T1 seedling roots formed callus-like tissue (Figure 6n,o). Upon transfer to soil, T1 plants formed multiple shoots reaching about a quarter of the height of the wild type (Figure 6a). Arabidopsis plants transformed with the same JuBel2 construct also formed an increased number of shoots (data not shown). Leaves of transgenic T1 tobacco plants resembled, in some features, phenotypes induced by the overexpression of class I Knox genes (Chan et al., 1998). Leaf veins diverged from the midrib at the base of the blade (Figure 6b), whereas ectopic organogenesis was never observed on leaves. Petioles lacked blade tissue at their flanks and formed a bulge of midrib tissue on their adaxial sides (Figure 6b,c). In the flowers of transgenic T0 and T1 plants ectopic outgrowths appeared on the outer surfaces of the petals. These appendages retained the identity of their subtending organs, to which they were fused at their bases (Figure 6f,h,i). Outgrowths terminating in sepal-like organs were also observed on the outside of the flower base (Figure 6e,h,i).

Figure 6.

Phenotypes of JuBel2 overexpressing tobacco plants.

(a) Two CaMV 35S JuBel2-transgenic T1 plants in comparison with non-transgenic tobacco (left).

(b,c) Leaves of non-transgenic (left) and transgenic (right) tobacco plants (b) and close-up of a transgenic petiole (c). Arrows indicate the petioles and the bulge on the transgenic petiole.

(d,e) Transgenic flower buds with appendages on the outsides of the petals and the flower base (e) in comparison to a non-transgenic bud (d).

(f) Transgenic flowers after removal of the sepals.

(g–i) Histological sections of a non-transgenic flower bud (g) compared to transgenic buds (h,i ).

(j,k) Wholemount preparations of non-transgenic (j) and transgenic (k) seedlings grown in sterile culture. Callus formation is indicated by arrows. fb: flower base; pa: petal appendages; pe: petal; sa: sepal appendages; se: sepal.


Our results show that the barley BKn-3 gene product interacts with proteins encoded by class I and II Knox genes. The three KNOX proteins form homodimers and heterodimerise in all possible combinations. JUBEL1 and JUBEL2 were identified as novel interacting partners of BKN3. They are encoded by members of the Knox-related Bell family of TALE homeobox genes. We suspect the existence of further BKN3-interacting proteins, as the failure to reisolate the already known partners BKN1, BKN3 and BKN7 indicated that our two-hybrid screen was not saturated. The BKN- and JUBEL-interactions were confirmed by in vitro experiments and two-hybrid interaction assays with deleted derivatives of the proteins. These assays were further aimed at determining their interaction requirements qualitatively. Steric hindrances due to the folding of the artificial chimeric proteins probably limit the interpretation of these results to some extent. In spite of these reservations, the following conclusions can be drawn: BKN3–KNOX and BKN3–JUBEL interactions occur independently of the N-terminus of BKN3, including the first one-third of the KNOX domain, while for BKN3–JUBEL interactions also the region C-terminal of the GSE-box is dispensable. The shortest JUBEL1 and JUBEL2 constructs interacting with BKN1 and BKN3 contain the putative coiled-coil domain and surrounding residues, supporting a role of this conserved region in mediating protein–protein interactions. JUBEL1 and JUBEL2 do not associate. Variations in the structural requirements of BKN3 for interactions with KNOX and JUBEL proteins suggest that similar but not identical interaction mechanisms are employed, which might be of considerable importance in the context of higher-order multiprotein complexes.

In situ hybridisation studies revealed that BKn-3 is expressed in meristematic tissues, in lateral organ primordia, young vegetative and reproductive organs and in tissues associated with the vasculature of the plant. BKn-1, BKn-7, JuBel1 and JuBel2 mRNAs show consistent spatial and temporal overlaps with BKn-3 expression. All in situ hybridisations were confirmed several times and in most cases with two gene-specific probes, allowing the conclusion that interactions amongst KNOX proteins and between KNOX and JUBEL proteins are topologically possible in planta. However, these in situ results provoke a critical reconsideration of the regulation of Knox gene function in SAM maintenance and leaf primordia initiation. The current model, proposed by Jackson et al. (1994), is based on the finding that Knox genes from different mono- and dicotyledonous plant species group into two meristematic expression categories. One pattern of class I Knox expression – typical for Kn-1 and Knox-8 of maize (Jackson et al., 1994), Osh-1 of rice (Sentoku et al., 1999), STM1 and ATK1/KNAT2 of Arabidopsis (Dockx et al., 1995; Long et al., 1996; Reiser et al., 2000) and Nth15 of tobacco (Tamaoki et al., 1997) – is characterised by signals throughout the SAM, while transcripts are excluded from lateral organs and their primordia. In the second category, mRNAs occupy positions intermediate between lateral organ primordia; this is typical for Rs-1, Knox-3 and Knox-4 of maize (Jackson et al., 1994; Reiser et al., 2000) and KNAT-1 of Arabidopsis (Lincoln et al., 1994). The model of Jackson et al. (1994) proposes that the domains of transcriptional activity of different class I Knox genes are responsible for the recurrent pattern of leaf-internode-leaf distribution during organ initiation.

Conflicting with this model, the tomato genes Tkn-1 and Tkn2/LeT6 are expressed in incipient and immature leaves and leaflet primordia, as well as in the SAM (Hareven et al., 1996; Janssen et al., 1998). The presence of class I Knox transcripts in leaf primordia was justified as a contribution of Knox gene products to a partial maintenance of meristematic identity, required for the formation of compound leaves. Although barley has simple leaves, we observed that two barley class I Knox genes are expressed in a similar pattern as the tomato genes. As also discussed by Parnis et al. (1997) in respect of the tomato genes, these findings contradict the assumption that the transcriptional down-regulation of class I Knox genes is the only prerequisite for organ initiation and differentiation at the flanks of the SAM. As a matter of fact, when ectopic meristems originating on the leaf surface of transgenic tobacco plants that expressed BKn-3 under control of the constitutive CaMV 35S promoter were hybridised with a BKn-3-specific probe, the presence of the transgene mRNA was restricted to these adventitious meristems. This suggested the involvement of mRNA accumulation and/or stability in the regulation of Knox gene function (K. J. Müller, unpublished results). Although we detect the BKn-3 mRNA all along the immature awn, immunolocalisation studies in Hooded barley with a maize KN1-specific antibody, also recognising BKN3, yielded signals in the lemma/awn transition zone only (Williams-Carrier et al., 1997). These findings indicate that post-transcriptional mechanisms should be considered in the discussion about the functional regulation of class I Knox genes.

JuBel2 was overexpressed in transgenic tobacco under control of the CaMV 35S promoter. Although N-terminally truncated, the construct used for transformation included all conserved domains of JUBEL2. Characteristics of the transformant's phenotypes were their bushy growth habit, the divergence of leaf veins at the leaf base and the appearance of ectopic outgrowths on the abaxial sides of both the petals and the flower base. Given the observation that it is the adaxial side of leaves that has the inherent capacity to maintain or regain meristematic and organogenic potential (Chan et al., 1998; McConnell and Barton, 1998; Waites et al., 1998), this characteristic of the 35S JuBel2 transformants is intriguing. Further, the effect of the transgene comes into action when sepal and petal primordia are already committed to their fate, indicating that homeobox genes can influence morphogenesis after MADS-box genes have conferred a specific identity to floral organs (Theißen and Saedler, 1999). In this sense, the JuBel2-induced floral phenotype is unique in flowering plants.

Also the possible role of KNOX proteins has been interpreted based on phenotypes resulting from their ectopic expression in transgenic model plants (Chan et al., 1998). In Arabidopsis, studies with different chimeric homeodomain polypeptides, assembled from class I and class II KNOX proteins, show that sequences outside the homeodomain determine gene-specific phenotypes (Serikawa and Zambryski, 1997). Similar experiments in tobacco confirm that the N-terminal half of KNOX proteins has functions in addition to transcriptional activation, possibly by mediating protein–protein interactions (Sakamoto et al., 1999). This interpretation converges with the results of our two-hybrid experiments.

Studies on vertebrate PBX and MEIS proteins, both of which belong to the TALE superclass of homeodomain proteins, strengthen the significance of KNOX–JUBEL interactions in plants: besides interacting with classical homeodomain proteins (like the HOX proteins), MEIS proteins associate with PBX proteins via the C-terminal half of their MEINOX domains (Mann and Affolter, 1998). Correspondingly, we have demonstrated that the C-terminal two-thirds of the KNOX (or MEINOX) domain of BKN3 are necessary for KNOX–KNOX and KNOX–JUBEL interactions. These interactions fulfil part of the prediction by Bürglin (1998), that interactions amongst TALE proteins and interactions between TALE proteins and classical homeodomain proteins should also exist in plants. While associations amongst classical plant homeodomain proteins have been described previously (Frank et al., 1998; Gonzalez et al., 1997; Sessa et al., 1993), data on their interactions with members of the TALE superfamily are not yet available. Still it can be taken for granted that the functional properties of plant homeodomain proteins are, similar to those of their animal counterparts, modulated by complex protein–protein interactions.

Experimental procedures

Construction of a two-hybrid cDNA expression library

Poly(A)+ mRNA was purified from total barley RNA prepared from developing inflorescences (Logemann et al., 1987) by adsorption to oligo-dT cellulose. cDNA was synthesised in the presence of methylated dCTP with the primer 5′-AGA TCT CGA GNN NNN N-3′, size-fractionated on a Sephacel column, and fragments of 0.3–2 kb were cloned into the EcoRI and XhoI digested arms of the HybriZAP vector (Stratagene, La Jolla, CA, USA). Packaging, amplification and in vivo excision of the pAD-GAL4 phagemid were carried out according to the vector kit. The primary library had a complexity of 2 × 106 Pfu with an average insert size of 400–800 bp.

Yeast two–hybrid interaction assays and two-hybrid screen

The vectors pACT2 and pAS2 (Clontech, Palo Alto, CA, USA) and pAD-GAL4 and pBD-GAL4 Cam (Stratagene) were used in the yeast strain Y190 (Harper et al., 1993). Plasmids were transformed by the LiOAc method (MATCHMAKER two-hybrid system, Clontech). β-Galactosidase assays followed Essers and Kunze (1996). Expression of the bait fusion protein in yeast was confirmed by Western analysis with a monoclonal antibody against the GAL4 DNA-binding domain (Clontech). Bait and library plasmids were sequentially transformed into Y190 and a total of 106 transformed cells was screened on medium lacking histidine, leucine and tryptophane and including 30 mm 3-aminotriazol (3AT). Growing colonies were tested for β-galactosidase activity. AD-plasmids were isolated according to the HybriZAPTM manual (Stratagene). Plasmid DNA was retransformed into Y190 and Y187 (Harper et al., 1993) in the presence or absence of the appropriate or of an inappropriate bait. Plasmids that induced both reporter gene expression only in the presence of the appropriate bait were sequenced.

AD and BD fusions were cloned by PCR using Pfu DNA polymerase (Stratagene). cDNA-fragments corresponding to the amino acid residues reported in Figure 3 were obtained with PCR primers including appropriate restriction sites. Fragments were cloned into the corresponding sites of the two-hybrid vector in frame with the respective GAL4 domain. All clones were verified by sequencing.

In vitro protein-binding assays

The coding regions of BKn-1, BKn-3 and BKn-7 and the cDNA fragment encoding amino acids 185–760 of JuBel1 were cloned into pGEX-5 × 1 (Pharmacia, Freiburg, Germany) in frame with the ORF for glutathione-S-transferase (GST) and used for purification of GST fusion proteins from E. coli BL21DE (Ausubel et al., 1994; Frangioni and Neel, 1993). cDNA fragments corresponding to amino acids 1–364 of BKN3, 329–760 of JUBEL1 and 122–341 of JUBEL2 were cloned into pBluescript KS and used as templates for in vitro transcription with T3 or T7 RNA polymerase (Roche, Mannheim, Germany). After in vitro translation in wheat germ extract (Promega, Madison, WI, USA) in the presence of [35S]methionine, equal amounts of 35S-labelled protein were incubated for 2 h at 4°C with glutathione-Sepharose-bound GST, GST fusion proteins (approximately 5 µg of each protein), and with the matrix alone in 200 µl of buffer (20 mm Tris–HCl pH 8, 100 mm NaCl, 2 mm EDTA, 0.1% (v/v) non-idet P-40). Bound proteins were eluted with 6% SDS-loading buffer and separated on 10% polyacrylamide gels under denaturing conditions (Laemmli, 1970). 35S-labelled proteins were detected by autoradiography after fixation (40% methanol, 10% acetate) and subsequent incubations of the gels in 10% DMSO, RotifluoreszintD (Roth, Karlsruhe, Germany) and water and drying them in a vacuum dryer.

Genomic and cDNA library screening and 5′-RACE amplifications

cDNA (cv. K-Atlas) and genomic (barley mutant CalC15) barley libraries (Müller, 1993) were screened according to Benton and Davis (1977). For cDNA-screening the JuBel1 and JuBel2 cDNA fragments isolated by two-hybrid screening were used as probes. cDNA inserts were cloned into pBluescript KS and sequenced. Genomic screening was carried out with intron probes amplified from barley genomic DNA. For JuBel1, the primer combinations JM117 (5′-GGG ACA GGC GCC ACG CCA TGC CGG-3′) × JM119 (5′-TGC ACA AGC AGC AGC ATG CAG TGG CAA G-3′) and JM110 (5′-CCA CAC ACA CGC ATT TGG GCT ATG-3′) × JM113 (5′-TGG CGT AGC ATG TGT CGC TGG CAC-3′) were used. JuBel2 probes were prepared with primer pairs JM121 (5′-TCG CCC CAC TAA GCA CAC CTT TTC CGA-3′) × JM122 (5′-GAG AGA GGT GAT TAA TTG CAG CCG GAT TAG-3′) and JM123 (5′-TGC TAA CAG TCT AGG TCC TCC GTT GCA TC-3′) × JM124 (5′-ATT ATT GTG CTG GTG CCT GTC ATG GTT TG-3′). Inserts of hybridising clones were subcloned in pBluescript KS and sequenced.

5′-RACE experiments were carried out with the SMARTTM RACE kit (Clontech) with cDNA prepared from young inflorescences of the varieties Bonus and Atlas (RNeasyTM Plant Mini Kit, Quiagen, Hilden, Germany) with the JuBel1-specific primers JM92 (5′-GCC GTC TCT CAC CCT TAG CTC CTC GGC-3′) and JM91 (5′-AGG GGA TGT CGG CGA CGA CCT TGA CTG C-3′, nested), and the JuBel2-specific primers JM94 (5′-GAG CAC CGA GGC GTA GCC GGT GAA GGG-3′) and JM93 (5′-CGA GAA GTT GTG CGT GGG GAT GTG CGG G-3′, nested). The amplification products were cloned with the Original TA Cloning Kit (Invitrogen, Karlsruhe, Germany) and sequenced.

Mapping of JuBel2

An intron fragment of 670 bp was amplified by PCR from genomic DNA of 100 lines of the barley Proctor × Nudinka mapping population (Heun et al., 1991), using primers JM123 (5′-TGC TAA CAG TCT AGG TCC TCC GTT GCA TC-3′) and JM124 (5′-ATT ATT GTG CTG GTG CCT GTC ATG GTT TG-3′). SNP heteroduplex analysis was carried out according to FMC BioProducts (FMC BioProducts, Rockland, ME, USA). For the mapping of JuBel2 to the linkage map of Castiglioni et al. (1998) the MAPMAKER/EXP 3.0b software (Lander et al., 1987) was utilised.

Sequencing, sequence analysis and primers

DNA sequences were determined by using the MPIZ DNA Core Facility on Applied Biosystems (Weiterstadt, Germany) Abi Prism 377 and 3700 sequencers. Sequence analysis was carried out with the Wisconsin Package Version 10.0, Genetics Computer Group (GCG, Madison, WI, USA). Sequence similarity searches were carried out with the NCBI's tool BLAST (Altschul et al., 1997). Oligonucleotides were purchased from Metabion (Planegg-Martinsried, Germany), MWG (Ebersberg, Germany) and Life Technologies (Karlsruhe, Germany).

Tobacco transformation

The JuBel2 construct was obtained by PCR amplification from pKS-JUBEL2(71–608) with a 5′ primer including an introduced ATG start-codon (5′-AAT AAA CTC GAG AAA ATG GAA TTC CGT CAC CAG CAC CAC CAG-3′) and integrated in pRT101 (Töpfer et al., 1987). The cassette including the CaMV 35S promoter, the JuBel2 cDNA and the polyadenylation signal was excised with HindIII and integrated into pBIN19 (Bevan, 1984). This construct was introduced into Agrobacterium tumefaciens LBA4404 (Hoekema et al., 1983) by electroporation and used for leaf disc transformation of the tobacco variety SRI (Matsuoka and Sanada, 1991).

Phenotypic analysis

For whole-mount preparations tobacco seedlings were fixed overnight in ethanol: acetic acid (6 : 1) at room temperature. After several washes in ethanol, they were mounted in chloralhydrate: glycerol: water (8 : 1 : 2), cleared for 8 h at room temperature and photographed with a Wild-Leitz Photoautomat MPS46. For histological sectioning the fixation, embedding, sectioning and dewaxing of tobacco flowers was carried out as described for the preparation of plant tissues for in situ hybridisation (Schmitz et al., 2000). After rehydration, sections were stained in 0.05% Toluidine blue for 2 min, washed in water and dehydrated.

In situ hybridisation

The BKn-1 probe spanned 740 bp 3′ to the ATG; the BKn-3 probe included the 5′ 330 bp of the cDNA. One BKn-7 probe included 107 bp upstream and 383 bp downstream of the ATG; the second probe started 30 bp upstream of the TGA and included 265 bp of the 3′-untranslated region. For both JuBel1 and JuBel2, in situ hybridisations were carried out with two gene-specific probes, corresponding to nucleotides 738–1563 and 2058–2819 for JuBel1, and 379–1083 and 1346–2080 for JuBel2. The specificity of all probes was verified by Southern hybridisation of genomic barley DNA. Probe templates were obtained by PCR and cloned into pBluescript KS. Plasmids were linearised and used for in vitro transcription of Digoxigenin (DIG)-labelled antisense RNAs.

Contamination of sense-probes with antisense transcripts during probe synthesis has been demonstrated by Northern hybridisation with single-stranded DNA-probes (J. Schmitz, personal communication). As a consequence, lambda-probes were used as alternative negative controls. As an additional confirmation, histone gene-(ubiquitous expression) and MADS-box gene-(tissue-specific expression patterns) specific probes were included as positive controls. For an additional negative control the labelled probe was omitted on one slide.

In situ hybridisations were carried out as described in Schmitz et al. (2000). As fixatives either 2% or 4% FAE (2% or 4% formaldehyde, 10% acetic acid, 50% ethanol) were used for 2–16 h at 4°C. Proteinase K (10–20 g ml−1) digestion was carried out at 37°C for 5–10 min in 2 × SSC, 0.1% SDS. Tissue sections were hybridised for 16 h at 50–55°C with 0.5–1 g l−1 probe concentration in a solution containing 50% formamide, 20 g l−1 tRNA, 10 ng l−1 Poly(A) RNA, 30 mm NaCl, 2 × TE pH 7.0, 1 × Denhardt's solution and 10% w/v dextran sulfate.


This work was in part supported by DFG grant RO330/8–1 and by BMBF grant 0311378 to W.R.

GenBank database accession numbers: JuBel1 AF334758, JuBel2 AF334759.